HW 01

Author

Amit Chawla

0 - Setup

if (!require("pacman")) 
  install.packages("pacman")
Loading required package: pacman
# use this line for installing/loading
# pacman::p_load() 

# devtools::install_github("tidyverse/dsbox")

library(tidyverse)
── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
✔ dplyr     1.1.4     ✔ readr     2.1.5
✔ forcats   1.0.0     ✔ stringr   1.5.1
✔ ggplot2   3.5.2     ✔ tibble    3.2.1
✔ lubridate 1.9.4     ✔ tidyr     1.3.1
✔ purrr     1.0.4     
── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag()    masks stats::lag()
ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
library(here)
here() starts at C:/Users/amitc/Home/MSDS-UofA/INFO-526/HW-01/hw-01-Amit9277
library(countdown)
library(scico) # for color palette 
library(DT) # for interactive table
library(readr)
library(ggmap)
ℹ Google's Terms of Service: <https://mapsplatform.google.com>
  Stadia Maps' Terms of Service: <https://stadiamaps.com/terms-of-service>
  OpenStreetMap's Tile Usage Policy: <https://operations.osmfoundation.org/policies/tiles>
ℹ Please cite ggmap if you use it! Use `citation("ggmap")` for details.
library(ggplot2)
library(ggrepel)
library(gridExtra)

Attaching package: 'gridExtra'

The following object is masked from 'package:dplyr':

    combine
library(dplyr)
library(pander)
library(lubridate)
library(hms)

Attaching package: 'hms'

The following object is masked from 'package:lubridate':

    hms
library(openintro)
Loading required package: airports
Loading required package: cherryblossom
Loading required package: usdata
# set theme for ggplot2
ggplot2::theme_set(ggplot2::theme_minimal(base_size = 14))

# set width of code output
options(width = 65)

# set figure parameters for knitr
knitr::opts_chunk$set(
  fig.width = 7,        # 7" width
  fig.asp = 0.618,      # the golden ratio
  fig.retina = 3,       # dpi multiplier for displaying HTML output on retina
  fig.align = "center", # center align figures
  dpi = 300             # higher dpi, sharper image
)

1 - Road traffic accidents in Edinburgh

accidents <- read_csv("data/accidents.csv")
Rows: 768 Columns: 31
── Column specification ─────────────────────────────────────────
Delimiter: ","
chr  (18): id, severity, date, day_of_week, highway, first_ro...
dbl  (12): easting, northing, longitude, latitude, police_for...
time  (1): time

ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.
# head(accidents)
# summary(accidents)
accidents |>
  mutate(
    time = as_hms(time),
    day_type = if_else(day_of_week %in% c("Saturday", "Sunday"), "Weekend", "Weekday")
  ) |>
  ggplot(aes(x = time, fill = severity)) +
  geom_density(alpha = 0.6, adjust = 1.2) +
  facet_wrap(~day_type, ncol = 1) +
  scale_fill_manual(values = c("Fatal" = "#8c6bb1", "Serious" = "#66c2a5", "Slight" = "#ffd92f")) +
  labs(
    title = "Number of accidents throughout the day",
    subtitle = "By day of week and severity",
    x = "Time of day",
    y = "Density",
    fill = "Severity"
  ) +
  theme_minimal(base_size = 14)

2 - NYC marathon winners

A.

nyc_marathon <- read_csv("data/nyc_marathon.csv")
Rows: 102 Columns: 7
── Column specification ─────────────────────────────────────────
Delimiter: ","
chr  (4): name, country, division, note
dbl  (2): year, time_hrs
time (1): time

ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.
# head(nyc_marathon)
# summary(nyc_marathon)
# Remove NA values
data <- na.omit(nyc_marathon)

# Histogram
ggplot(data, aes(x = time_hrs)) +
  geom_histogram(binwidth = 0.1, fill = "skyblue", color = "black") +
  labs(title = "Histogram of Marathon Times", x = "Time (hours)", y = "Frequency")

# Box plot
ggplot(data, aes(y = time_hrs)) +
  geom_boxplot(fill = "lightgreen") +
  labs(title = "Box Plot of Marathon Times", y = "Time (hours)")

Histogram insights (not in box plot):

  • Shows distribution shape (e.g., slight skewness).

  • Can detect multimodality or clusters.

  • Visualizes frequency and spread more granularly.

Box plot insights (not in histogram):

  • Summarizes five-number summary (min, Q1, median, Q3, max).

  • Shows outliers more clearly.

  • Easier to compare central tendency and spread at a glance.

B.

ggplot(data, aes(x = division, y = time_hrs, fill = division)) +
  geom_boxplot() +
  scale_fill_manual(values = c("Men" = "steelblue", "Women" = "coral")) +
  labs(title = "Marathon Times by Division", x = "Division", y = "Time (hours)") +
  theme_minimal()

Comparison:

  • Women generally have longer marathon times than men.

  • Both distributions show similar variability, but the center for women is shifted right.

  • Some women’s results have more outliers, especially earlier years when participation was lower.

C.

In part b, both x = division and fill = division were used — this is redundant, because the x-axis already separates the divisions.

ggplot(data, aes(x = division, y = time_hrs)) +
  geom_boxplot(fill = c("steelblue", "coral")) +
  labs(title = "Marathon Times by Division", x = "Division", y = "Time (hours)") +
  theme_minimal()

Effect:

  • Reduces redundancy in encoding the same info with both x-position and color.

  • Clean visual, higher data-to-ink ratio (i.e., less decorative ink for the same amount of information).

D.

ggplot(data, aes(x = year, y = time_hrs, color = division, shape = division)) +
  geom_point(size = 3) +
  scale_color_manual(values = c("Men" = "steelblue", "Women" = "coral")) +
  labs(title = "NYC Marathon Times Over Years", x = "Year", y = "Time (hours)") +
  theme_minimal()

Insights only visible in this plot:

  • Clear improvement over time (downward trend in times for both genders).

  • Men and women’s times converged somewhat over decades.

  • Irregularities in early years (especially for women) show inconsistency in participation or race organization.

  • Gaps or spikes might highlight historic events (e.g., no female finisher in 1970, or race disruptions).

3 - US counties

library(usdata)

A.

ggplot(county) +
  geom_point(aes(x = median_edu, y = median_hh_income)) +
  geom_boxplot(aes(x = smoking_ban, y = pop2017))
Warning: Removed 3 rows containing non-finite outside the scale range
(`stat_boxplot()`).
Warning: Removed 2 rows containing missing values or values outside the
scale range (`geom_point()`).

*Does it work?**

No, this code will likely produce an error or unexpected results.

*Why doesn’t it make sense?**

  • The first layer (geom_point) is fine: it creates a scatterplot of median education (years of schooling) vs median household income. That works.

  • The second layer (geom_boxplot) is the problem. Boxplots require a categorical variable on the x-axis and a numeric variable on the y-axis. While smoking_ban is categorical (e.g., “Yes”, “No”, or NA), pop2017 is numeric and suitable.

    However, the problem is plotting two incompatible geometries (scatterplot and boxplot) on the same plot without separating them, which can cause:

    • Misleading or unreadable visuals

    • Warnings or errors due to mismatch in the x/y aesthetics between layers

B.

# First plot (facets stacked vertically)
ggplot(county %>% filter(!is.na(median_edu))) +
  geom_point(aes(x = homeownership, y = poverty)) +
  facet_grid(median_edu ~ .)

# Second plot (facets side by side)
ggplot(county %>% filter(!is.na(median_edu))) +
  geom_point(aes(x = homeownership, y = poverty)) +
  facet_grid(. ~ median_edu)

*Which one is better?**

The second plot (facet_grid(. ~ median_edu)) is better for comparing poverty levels across different median education levels.

Why?

  • Comparing across columns (i.e., horizontal layout) makes it easier for the eye to scan vertically and compare trends in poverty across education levels.

  • If education is the main grouping variable, placing it across columns makes patterns more obvious.

  • Plotting education levels side-by-side allows you to compare similar homeownership vs poverty relationships across education levels more naturally.

Takeaway on faceting:

  • Use facet_grid(. ~ variable) (columns) when you want to compare groups side by side.

  • Use facet_grid(variable ~ .) (rows) when vertical comparison is more appropriate, or you have long text labels that don’t fit well in horizontal space.

C.

Plot A – Basic scatter plot

ggplot(data = county, aes(x = homeownership, y = poverty)) +
  geom_point() +
  ggtitle("Plot A")
Warning: Removed 2 rows containing missing values or values outside the
scale range (`geom_point()`).

Plot B – Scatter plot with a smooth (loess) trend line

ggplot(data = county, aes(x = homeownership, y = poverty)) +
  geom_point() +
  geom_smooth(se = FALSE) +
  ggtitle("Plot B")
`geom_smooth()` using method = 'gam' and formula = 'y ~ s(x, bs
= "cs")'
Warning: Removed 2 rows containing non-finite outside the scale range
(`stat_smooth()`).
Warning: Removed 2 rows containing missing values or values outside the
scale range (`geom_point()`).

Plot C – Scatter plot with a loess smoother in green including confidence intervals

ggplot(data = county, aes(x = homeownership, y = poverty)) +
  geom_point() +
  geom_smooth(aes(group = metro), se = FALSE, color = "green") +
  ggtitle("Plot C")
`geom_smooth()` using method = 'gam' and formula = 'y ~ s(x, bs
= "cs")'
Warning: Removed 2 rows containing non-finite outside the scale range
(`stat_smooth()`).
Warning: Removed 2 rows containing missing values or values outside the
scale range (`geom_point()`).

Plot D – Scatter plot with smoother and no confidence intervals

ggplot(data = county, aes(x = homeownership, y = poverty)) +
  geom_smooth(aes(group = metro), se = FALSE) +
  geom_point() +
  ggtitle("Plot D")
`geom_smooth()` using method = 'gam' and formula = 'y ~ s(x, bs
= "cs")'
Warning: Removed 2 rows containing non-finite outside the scale range
(`stat_smooth()`).
Warning: Removed 2 rows containing missing values or values outside the
scale range (`geom_point()`).

Plot E – Scatter plot colored by metro, with separate lines for each metro level using linetype

ggplot(data = county, aes(x = homeownership, y = poverty, color = metro)) +
  geom_point() +
  geom_smooth(aes(linetype = metro), se = FALSE, color = "blue") +
  ggtitle("Plot E")
`geom_smooth()` using method = 'gam' and formula = 'y ~ s(x, bs
= "cs")'
Warning: Removed 2 rows containing non-finite outside the scale range
(`stat_smooth()`).
Warning: Removed 2 rows containing missing values or values outside the
scale range (`geom_point()`).

Plot F – Scatter plot with colored smoothers for each level of metro

ggplot(data = county, aes(x = homeownership, y = poverty, color = metro)) +
  geom_point() +
  geom_smooth(se = FALSE) +
  ggtitle("Plot F")
`geom_smooth()` using method = 'gam' and formula = 'y ~ s(x, bs
= "cs")'
Warning: Removed 2 rows containing non-finite outside the scale range
(`stat_smooth()`).
Warning: Removed 2 rows containing missing values or values outside the
scale range (`geom_point()`).

Plot G – Scatter plot with colored points and a single smoother line across all data

ggplot(data = county, aes(x = homeownership, y = poverty, color = metro)) +
  geom_point() +
  geom_smooth(aes(group = 1), se = FALSE, color = "blue") +
  ggtitle("Plot G")
`geom_smooth()` using method = 'gam' and formula = 'y ~ s(x, bs
= "cs")'
Warning: Removed 2 rows containing non-finite outside the scale range
(`stat_smooth()`).
Warning: Removed 2 rows containing missing values or values outside the
scale range (`geom_point()`).

Plot H – Colored scatter plot without any smoothers

ggplot(data = county, aes(x = homeownership, y = poverty, color = metro)) +
  geom_point() +
  ggtitle("Plot H")
Warning: Removed 2 rows containing missing values or values outside the
scale range (`geom_point()`).

4 - Rental apartments in SF

credit <- read_csv("data/credit.csv")
Rows: 400 Columns: 5
── Column specification ─────────────────────────────────────────
Delimiter: ","
chr (2): student, married
dbl (3): balance, income, limit

ℹ Use `spec()` to retrieve the full column specification for this data.
ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.
head(credit)
# A tibble: 6 × 5
  balance income student married limit
    <dbl>  <dbl> <chr>   <chr>   <dbl>
1     333   14.9 No      Yes      3606
2     903  106.  Yes     Yes      6645
3     580  105.  No      No       7075
4     964  149.  No      No       9504
5     331   55.9 No      Yes      4897
6    1151   80.2 No      No       8047
summary(credit)
    balance            income         student         
 Min.   :   0.00   Min.   : 10.35   Length:400        
 1st Qu.:  68.75   1st Qu.: 21.01   Class :character  
 Median : 459.50   Median : 33.12   Mode  :character  
 Mean   : 520.01   Mean   : 45.22                     
 3rd Qu.: 863.00   3rd Qu.: 57.47                     
 Max.   :1999.00   Max.   :186.63                     
   married              limit      
 Length:400         Min.   :  855  
 Class :character   1st Qu.: 3088  
 Mode  :character   Median : 4622  
                    Mean   : 4736  
                    3rd Qu.: 5873  
                    Max.   :13913

A. Relationship between income and credit card balance:

Non-students (blue): Slight positive trend; balance increases from $0 to $1,500–$2,000 as income rises from $50K to $150K.

Students (red): Stronger positive trend; balances start higher ($500–$1,500 at $50K) and increase more with income, especially for unmarried students.

Marital status impact: Unmarried students show a steeper balance increase with income; married students have flatter trends.

# Create the plot
ggplot(data = credit, aes(x = income, y = balance, color = student, shape = student)) +
  geom_point(alpha = 0.5) +  # Add transparency to points (faded)
  geom_smooth(method = "lm", se = FALSE) +  # Add regression line
  facet_grid(student ~ married, 
             labeller = labeller(student = c("No" = "Student: No", "Yes" = "Student: Yes"),
                                 married = c("No" = "Married: No", "Yes" = "Married: Yes"))) +  # Custom facet labels
  scale_color_manual(values = c("No" = "#1f77b4", "Yes" = "#ff7f0e")) +  # Blue for No, orange for Yes
  scale_shape_manual(values = c("No" = 16, "Yes" = 17)) +  # 16 = circle, 17 = triangle
  labs(x = "Income", y = "Credit card balance") +
  scale_x_continuous(labels = function(x) paste0("$", x, "K"),  # Add $ and K for income
                     limits = c(0, 150), breaks = seq(0, 150, by = 50)) +
  scale_y_continuous(labels = function(y) paste0("$", y),  # Add $ for balance
                     limits = c(0, 2000), breaks = seq(0, 2000, by = 500)) +
  theme_minimal() +
  theme(legend.position = "none")  # Remove legend since colors are indicated by facets
`geom_smooth()` using formula = 'y ~ x'
Warning: Removed 9 rows containing non-finite outside the scale range
(`stat_smooth()`).
Warning: Removed 9 rows containing missing values or values outside the
scale range (`geom_point()`).

B.

  • Student Status: Yes, “student” is likely a useful predictor. Students consistently have higher credit card balances than non-students at the same income level, and their balances increase more rapidly with income. This suggests that being a student influences spending or credit behavior, possibly due to educational expenses or limited income sources leading to reliance on credit.
  • Marital Status: Yes, “married” is also likely a useful predictor, particularly for students. The relationship between income and balance for students differs significantly between married and unmarried individuals. Unmarried students show a steeper increase in balance with income, while married students have a more stable balance across income levels. For non-students, the effect of marital status is less pronounced, but there are still slight differences in the spread of balances.

C.

Credit utilization = balance/limit. The plot shows utilization vs. income, split by married/student status.

# Calculate credit utilization
credit <- credit %>%
  mutate(credit_utilization = balance / limit)

# Create subsets for each combination
ggplot(credit, aes(x = income, y = credit_utilization * 100)) +
  geom_point(aes(color = student, shape = student), alpha = 0.5) +
  geom_smooth(
    aes(group = interaction(student, married), color = student),
    method = "lm", se = FALSE
  ) +
  facet_grid(student ~ married,
             labeller = labeller(
               student = c("No" = "Student: No", "Yes" = "Student: Yes"),
               married = c("No" = "Married: No", "Yes" = "Married: Yes")
             )) +
  scale_color_manual(values = c("No" = "#1f77b4", "Yes" = "#ff7f0e")) +
  labs(x = "Income", y = "Credit utilization") +
  scale_x_continuous(
    labels = function(x) paste0("$", x, "K"),
    breaks = seq(0, 150, by = 50)
  ) +
  scale_y_continuous(
    labels = scales::percent_format(scale = 1),
    breaks = seq(0, 25, by = 10)
  ) +
  coord_cartesian(xlim = c(0, 150), ylim = c(0, 25)) +
  theme_minimal() +
  theme(legend.position = "right", legend.title = element_blank())
`geom_smooth()` using formula = 'y ~ x'

D. Differences in relationships (credit utilization vs. credit balance):

  • Credit Balance: Positive trend with income; steeper for students, especially unmarried ones.

  • Credit Utilization: Negative or flat trend with income. Non-students’ utilization slightly increases (0% to 10%); students’ utilization decreases (20% to 10%), with little difference by marital status.

  • Key Difference: Utilization decreases or stays stable with income, while balance increases, suggesting higher incomes lead to higher limits, reducing utilization. Student status still matters, but marital status has less impact on utilization trends.

5 - Napoleon’s march.

Introduction

Charles Joseph Minard’s famous 1869 flow map depicts Napoleon’s disastrous Russian campaign of 1812. This iconic graphic communicates multivariate data with impressive clarity—showing geography, temperature, troop size, and direction—all on one chart. In this exercise, I will recreate this visualization using ggplot2 in R. The data is structured into three data frames: troops, cities, and temperatures, available in the napoleon.rds file.

Resources Used

  1. Andrew Heiss’s blog post
    URL: Exploring Minard’s 1812 plot with ggplot2
    Usage: Used as the primary reference for understanding how to work with Minard’s structured dataset and layer paths, points, and text annotations in ggplot2.

  2. Hadley Wickham’s paper on the layered grammar of graphics
    URL: Layered Grammar of Graphics (PDF)
    Usage: Provided conceptual foundation for thinking in terms of layered components of ggplot2—how we add geoms, mappings, and data transformations step by step.

Code: Recreating the March with a Personal Touch

# Load required libraries
library(ggplot2)
library(ggrepel)
library(dplyr)
library(grid)
library(readr)
library(scales)  # For comma_format()

Attaching package: 'scales'
The following object is masked from 'package:purrr':

    discard
The following object is masked from 'package:readr':

    col_factor
# Load the napoleon.rds file (adjust path as needed)
napoleon <- read_rds("data/napoleon.rds")

# Unpack the dataset into separate data frames for clarity
troops <- napoleon$troops       # Troop movement data (long, lat, survivors, direction, group)
cities <- napoleon$cities       # City locations (long, lat, city)
temps <- napoleon$temperatures  # Temperature data (long, temp, month, day)

# Create a data frame for key historical events to annotate on the plot
# Placeholder coordinates for significant events; adjust if precise locations are known
events <- data.frame(
  long = c(37.5, 32.0),  # Longitude for Battle of Borodino and Crossing of Berezina
  lat = c(55.5, 54.0),   # Latitude for these events
  event_name = c("Battle of Borodino", "Crossing of Berezina")  # Event names to display
)

# Prepare data for survivor labels by selecting key points
# Match troop points to cities and include start/end points of each group
label_points <- troops %>%
  # Group by army corps (group) and direction (advance/retreat) to process each segment
  group_by(group, direction) %>%
  # Join with cities to find troop points matching city coordinates
  inner_join(cities, by = c("long", "lat"), suffix = c(".troops", ".cities")) %>%
  # Keep only necessary columns for labeling
  select(long, lat, survivors, direction, group, city) %>%
  ungroup() %>%
  # Add start and end points of each group/direction to capture initial and final survivor counts
  bind_rows(
    troops %>%
      group_by(direction, group) %>%
      slice(c(1, n())) %>%  # Select first and last points of each group
      ungroup() %>%
      distinct(long, lat, .keep_all = TRUE) %>%  # Remove duplicate coordinates
      select(long, lat, survivors, direction, group)
  ) %>%
  # Remove any duplicate points to avoid redundant labels
  distinct(long, lat, survivors, direction, group, .keep_all = TRUE)

# Create the main troop movement plot to visualize Napoleon's march
march.1812.plot.simple <- ggplot() +
  # Plot troop paths, with color indicating direction and linewidth showing survivor numbers
  geom_path(data = troops, aes(x = long, y = lat, group = group,
                               color = direction, linewidth = survivors),
            lineend = "round") +  # Use rounded line ends for smoother appearance
  # Add survivor labels at city points and start/end, in dark blue for visibility
  geom_text_repel(data = label_points, 
                  aes(x = long, y = lat, label = scales::comma(survivors)),
                  color = "#00008B", size = 2, family = "sans",
                  segment.color = "grey50", segment.size = 0.2,
                  nudge_y = 0.15, direction = "y",  # Nudge upward to avoid city labels
                  max.overlaps = 30, force = 2) +  # Increase repulsion to reduce overlap
  # Plot city points as red markers
  geom_point(data = cities, aes(x = long, y = lat),
             color = "#DC5B44") +
  # Add city name labels, nudged downward to separate from survivor labels
  geom_text_repel(data = cities, aes(x = long, y = lat, label = city),
                  color = "#DC5B44", size = 2, family = "sans",
                  nudge_y = -0.15, direction = "y",  # Nudge downward for clarity
                  max.overlaps = 30, force = 2) +
  # Annotate key historical events in bold dark red, positioned above other labels
  geom_text(data = events, aes(x = long, y = lat, label = event_name),
            nudge_y = 0.3, color = "darkred", size = 2, fontface = "bold") +
  # Set linewidth scale for survivors without a legend to keep plot clean
  scale_linewidth_continuous(range = c(0.5, 10), guide = "none") +
  # Add direction legend with gold for advance and black for retreat
  scale_color_manual(name = "Direction", values = c("#FFD700", "#000000"), 
                     labels = c("Advance", "Retreat")) +
  # Use minimal theme and place legend at bottom for accessibility
  theme_minimal() +
  theme(legend.position = "bottom")

# Prepare temperature data with formatted labels for the temperature plot
temps.nice <- temps %>%
  mutate(nice.label = paste0(temp, "°, ", month, ". ", day))  # Combine temp, month, day into a single label

# Create the temperature plot to show conditions during the retreat
temps.1812.plot <- ggplot(data = temps.nice, aes(x = long, y = temp)) +
  # Add a dashed red line at 0°C to highlight freezing conditions
  geom_hline(yintercept = 0, linetype = "dashed", color = "red") +
  # Plot temperature as a line over longitude
  geom_line() +
  # Add labels for temperature points with date information
  geom_label(aes(label = nice.label),
             family = "sans", size = 1.5) +  # Smaller size for less clutter
  # Set axis labels and scales, aligning x-axis with main plot
  labs(x = NULL, y = "° Celsius") +
  scale_x_continuous(limits = ggplot_build(march.1812.plot.simple)$layout$panel_ranges[[1]]$x.range) +
  scale_y_continuous(position = "right") +  # Place y-axis on right for clarity
  coord_cartesian(ylim = c(-35, 5)) +  # Set y-axis limits with padding
  # Use clean theme, removing unnecessary grid lines and x-axis elements
  theme_bw(base_family = "sans") +
  theme(panel.grid.major.x = element_blank(),
        panel.grid.minor.x = element_blank(),
        panel.grid.minor.y = element_blank(),
        axis.text.x = element_blank(), 
        axis.ticks = element_blank(),
        panel.border = element_blank())

# Combine the troop movement and temperature plots vertically
both.1812.plot.simple <- rbind(ggplotGrob(march.1812.plot.simple),
                               ggplotGrob(temps.1812.plot))

# Adjust panel heights to a 3:1 ratio for a wider appearance
panels <- both.1812.plot.simple$layout$t[grep("panel", both.1812.plot.simple$layout$name)]
both.1812.plot.simple$heights[panels] <- unit(c(3, 1), "null")

# Render the combined plot interactively
grid::grid.newpage()
grid::grid.draw(both.1812.plot.simple)

Code Explanation:

This R script generates a two-part visualization of Napoleon’s 1812 winter retreat from Moscow, inspired by Charles Minard’s iconic graphic. The code processes data, creates a troop movement plot and a temperature plot, and combines them to depict the army’s journey, survivor losses, and environmental conditions.

The script loads data from napoleon.rds, processes it to extract key points, and uses ggplot2 to create: - A troop movement plot showing paths with survivor counts, city locations, and historical event annotations. - A temperature plot showing temperature changes aligned with the troop paths. - A combined visualization stacking both plots vertically with a wider layout.

Visualization Approach

  1. Data Loading and Preparation:
    • Loads napoleon.rds and unpacks it into troops (path data with longitude, latitude, survivors, direction, group), cities (city locations), and temperatures (temperature data with dates).
    • Creates an events data frame for annotating historical events (Battle of Borodino, Crossing of Berezina) using placeholder coordinates.
    • Prepares survivor labels by matching troops coordinates to cities and including start/end points of each group/direction using dplyr. Duplicates are removed to ensure unique labels. The scales package formats survivor counts in normal notation (e.g., 422,000).
  2. Troop Movement Plot:
    • Uses ggplot2 to plot troop paths with geom_path, coloring advance (gold, #FFD700) and retreat (black, #000000), with linewidths scaled to survivor numbers (0.5–10 range).
    • Adds survivor labels in dark blue (#00008B) at city matches and start/end points using geom_text_repel with scales::comma for clarity. Labels are nudged upward (nudge_y = 0.15) to avoid overlap.
    • Marks cities with red (#DC5B44) points and labels, nudged downward (nudge_y = -0.15) for separation. Event annotations in bold dark red are placed higher (nudge_y = 0.3).
    • Includes a direction legend at the bottom, suppresses the survivors legend, and uses theme_minimal for a clean look.
  3. Temperature Plot:
    • Plots temperature over longitude with geom_line, adding a dashed red 0°C line (geom_hline) to highlight freezing conditions.
    • Labels temperature points with date and temperature (e.g., -30°, Oct. 25) using geom_label with a smaller size (1.5) for clarity.
    • Aligns the x-axis with the troop plot and sets the y-axis (right side) from -35°C to 5°C. Uses theme_bw with minimal grid lines for a clean appearance.
  4. Combining and Rendering:
    • Combines plots vertically using rbind and ggplotGrob, setting a 3:1 height ratio (unit(c(3, 1), "null")) for a wider troop plot.
    • Renders the combined plot interactively with grid::grid.draw for display in R environments like RStudio.

Result

The code produces a visualization showing Napoleon’s army shrinking along its path, with survivor counts labeled at key cities (e.g., Moscow, Smolensk) and endpoints, city names, and event annotations for context. The temperature plot below links harsh conditions to troop losses, creating a clear, informative graphic of the campaign’s toll.

What Have I add different from the article:

  • Added Survivor Labels on the Plot:

    • Change: Introduced a label_points data frame to select survivor counts at city matches and start/end points of each group/direction, using geom_text_repel with scales::comma(survivors) to display numbers in normal notation (e.g., 422,000) in dark blue (#00008B).

    • Reason: To show survivor numbers directly on the plot, addressing your request to display more data from the troops dataset and make the human cost of the campaign explicit. Dark blue was chosen for visibility, and geom_text_repel with nudge_y = 0.15 prevents overlap with other elements.

  • Added Event Annotations:

    • Change: Created an events data frame with placeholder coordinates for historical events (Battle of Borodino, Crossing of Berezina) and added geom_text to annotate these in dark red.

    • Reason: Enhances historical context by marking significant events, making the plot more informative about key moments in Napoleon’s campaign.

  • Added 0°C Reference Line in Temperature Plot:

    • Change: Added geom_hline(yintercept = 0, linetype = "dashed", color = "red") to the temperature plot.

    • Reason: Highlights freezing conditions that impacted the army, visually linking environmental factors to troop losses for a clearer narrative.

  • Modified City and Survivor Label Positioning:

    • Change: Adjusted geom_text_repel for city labels with nudge_y = -0.15 to push them downward and for survivor labels with nudge_y = 0.15 to push them upward, both with max.overlaps = 30 and force = 2. Reduced label sizes (size = 2 for survivor/city, 1.5 for temperature).

    • Reason: Prevents overlap between city names and survivor numbers, improving readability. Smaller label sizes reduce visual clutter while maintaining clarity.

  • Changed Color Scheme for Direction:

    • Change: Updated scale_colour_manual values from c(“#DFC17E”, “#252523”) to c(“#FFD700”, “#000000”) (gold and black) and added a direction legend with name = "Direction" and labels = c("Advance", "Retreat").

    • Reason: Gold and black provide higher contrast for advance and retreat paths, improving visual distinction. The legend clarifies direction, replacing the suppressed guides(color = "none").

  • Replaced scale_size with scale_linewidth_continuous:

    • Change: Changed scale_size(range = c(0.5, 10)) to scale_linewidth_continuous(range = c(0.5, 10), guide = "none").

    • Reason: scale_linewidth_continuous is the correct function for linewidth aesthetic in newer ggplot2 versions, ensuring proper scaling of troop path widths without a legend.

  • Changed Theme for Main Plot:

    • Change: Replaced theme_nothing() with theme_minimal() and added theme(legend.position = "bottom").

    • Reason: theme_minimal() provides a cleaner background with axes for context, while theme_nothing() was overly sparse. The bottom legend position keeps the plot uncluttered.

  • Reduced Label Sizes:

    • Change: Set size = 2 for survivor and city labels (from 3) and size = 1.5 for temperature labels (from 2.5).

    • Reason: Smaller sizes reduce visual clutter, especially with more labels added, while maintaining readability.